OpenAI Adds Zero Data Retention and Private Safety Processing for Enterprise AI

OpenAI has introduced two new capabilities aimed at enterprise AI deployments: Zero Data Retention (ZDR) and Private Safety Processing. These features are designed to address enterprise concerns around data privacy and safety monitoring for sensitive AI applications.

ZDR means that OpenAI will not retain customer prompts or model responses after a single request. For enterprise data, it will not be used for training unless the customer explicitly agrees. The data can either remain on customer-controlled infrastructure or be stored on OpenAI infrastructure using encryption keys controlled by the customer. This allows enterprises to keep their data private while still using OpenAI's frontier models.

Private Safety Processing is a new safety layer that extends monitoring across related interactions, rather than assessing each request individually. This broad approach enables pattern-based risk detection, even if individual requests do not reveal the full context of a potential issue. When a risk is identified, a narrowly defined signal is sent to OpenAI for enforcement decisions.

Importantly, the underlying customer content is not exposed to OpenAI personnel; instead, investigation and auditing data remains within the customer's system. This separation of content and enforcement signals can help maintain internal audit trails while limiting exposure of sensitive data.

Combined, these capabilities aim to provide enterprise customers with clearer boundaries around data handling for frontier model deployments. However, the specifics of which OpenAI products and models are covered, as well as pricing details, remain unclear. Enterprises should treat these announcements as significant capabilities but verify availability and contractual scope for their intended deployment.